利用不比例分析进行信号检测

Disproportionality analysis is a core data mining technique in modern pharmacovigilance. It addresses the challenge of finding a ‘needle in a haystack’ within massive spontaneous reporting databases containing millions of reports. The fundamental idea is to compare the proportion of reports for a specific adverse event with a specific drug to the proportion of reports for that same event with all other drugs in the database. If a drug-event pair appears significantly more often than would be expected by chance, it is flagged as a ‘signal’ of a potential association.

This is typically calculated using a 2×2 contingency table. For a given drug (Drug X) and event (Event Y), the table contains four cells: (a) reports with Drug X and Event Y, (b) reports with Drug X and any other event, (c) reports with any other drug and Event Y, and (d) reports with any other drug and any other event. The Reporting Odds Ratio (ROR) is then calculated as [latex](a/c) / (b/d) = ad/bc[/latex]. A ROR value significantly greater than 1, along with a sufficient number of cases, suggests a statistical association.

Other common measures include the Proportional Reporting Ratio (PRR) and Bayesian methods like the Multi-item Gamma Poisson Shrinker (MGPS). It is crucial to understand that these methods do not establish causality. They are hypothesis-generating tools. A statistical signal can be influenced by many 偏见, such as media attention (the ‘Weber effect’), co-prescribed medications, or the underlying disease being treated. Therefore, any detected signal must undergo a thorough qualitative and clinical assessment by experts before any regulatory action is considered.